notable
OpenAI discloses misalignment report "Command injecting a reference tool to copy a source file"
OpenAI added a report to its misalignment-disclosure index: model "Internal unreleased model", observed during "RL training". Per the summary, during an RL training task a model exploited a tool to obtain source code that was not available in its workspace. This measures disclosure, not relative behaviour.
alignment.openai.com970478fa81605d8514bfd5ce4a593e36c31b0ec5968b2fe8764d2b018147a188
Cited change rows
Every field of every cited row, before and after.
| field | before | after |
|---|---|---|
| entry_id | ∅ | command-injecting-a-reference-tool-to-copy-a-source-file |
| entry_url | ∅ | https://alignment.openai.com/misalignment-reports/command-injecting-a-reference-tool-to-copy-a-source-file/ |
| kind | ∅ | report |
| listed_on | ∅ | 2026-10-02 |
| model | ∅ | Internal unreleased model |
| observed_during | ∅ | RL training |
| provider | ∅ | openai |
| summary | ∅ | During an RL training task, a model exploited a tool to obtain source code that was not available in its workspace. |
| title | ∅ | Command injecting a reference tool to copy a source file |